Bound et al 2015 — Measuring Recent Apparent Declines In Longevity The Role Of Increasing Educational Attainment

Summary

Using the same Multiple Cause of Death data as Olshansky et al. (2012), Bound et al. replicate the widely publicized finding of dramatic life expectancy (LE) declines for non-Hispanic whites without a high school diploma, then show the finding nearly collapses when education is reclassified by relative rank within birth cohorts rather than by absolute credential. The core mechanism is compositional selection bias: US high school graduation rates rose sharply over the 20th century, so "less than high school" captured the bottom $55\%$ of the population in 1990 but only the bottom $37\%$ by 2008 — an increasingly adversely selected residual. Using bottom-quartile rank instead, the apparent $3.9$-year LE decline for white women shrinks to $1.2$ years; the apparent $2.2$-year decline for white men becomes a $0.4$-year gain. What survives the correction: widening LE inequality between top and bottom quartiles, genuine stagnation for white women of low socioeconomic status (SES), and entrenched black-white gaps.

Key Claims

• Replication (absolute credential): Using the Olshansky et al. approach, Bound et al. confirm white women with less than high-school (HS) education had $P(\text{survive to }85)$ decline from $40.9\%$ (1990) to $31.4\%$ (2010); white men from $17.6\%$ to $16.7\%$. Expected years lived 25–85 fell $3.9$ years for women and $2.2$ years for men. The finding replicates.
• Relative rank correction: Using bottom educational quartile of each birth cohort instead:
  • White women: LE declined modestly by $1.2$ years (not $3.9$); no appreciable change through age 65, some drop at older ages only
  • White men: LE increased $0.4$ years (not declined $2.2$ years); survival curves roughly flat through middle age
  • Black men and women: no decline on either measure
• The mechanism — compositional selection bias: HS graduation rates rose from $\approx 45\%$ (cohorts born 1903–1921) to $\approx 63\%$ (cohorts born 1919–1937). The "less than HS" group shrank from bottom $55\%$ of population (1990) to bottom $37\%$ (2008). The residual non-diploma group grows increasingly adversely selected over time, elevating their mortality by composition, not by causal change. Dowd and Hamoudi (2014) proved this formally via simulation: rising educational attainment alone can reproduce the Olshansky pattern even when every group's LE improves.
• What survives: Widening inequality is real — white women top 75% gained $1.2$ years, bottom quartile lost $1.2$ years; white men top 75% gained $2.8$ years. Black-white gaps are entrenched across all education levels. Low SES groups, especially white women over 65, face genuine stagnation.
• Behavioral explanations are overstated: Smoking differentials can account for only a small portion of growing LE gaps; obesity evidence is mixed. The authors emphasize social determinants — chronic stress, allostatic load ("weathering"), unequal access to medical advances — as the more plausible drivers of persistent low-SES mortality disadvantage.
• Social Security retirement age implication: Raising Social Security (SS) or Medicare eligibility ages based on aggregate LE gains is regressive — low-SES groups not sharing in those gains bear the full burden.

Concepts Introduced or Extended

• Income-Mortality Gradient
• Period Mortality

Entities Mentioned

• John Bound
• Arline Geronimus
• Hilary Waldron

Quotes

"Focusing on mortality rates of those with less than a high school education over a twenty-year period means looking at a different, shrinking, and increasingly vulnerable segment of the population in each year."

"Robust evidence refutes a dramatic decline in life expectancy for any race-sex group in the United States in recent decades…the more modest evidence of a decline for white women of low socioeconomic status and the entrenchment or even growth of other race-sex socioeconomic inequalities in longevity appear to signal chronic and insidious processes."

My Take

The paper's core methodological contribution is clean and the simulation evidence from Dowd and Hamoudi makes it nearly irrefutable: the Olshansky finding is largely an artifact of a changing denominator. The surviving $1.2$-year decline for white women — modest but real — is worth noting precisely because it is not alarming enough to generate front-page coverage, yet it documents genuine stagnation in an era of broad mortality improvement. The weathering/allostatic load framework (Geronimus's signature contribution) is less empirically developed in this paper than in her dedicated work, but it provides a mechanistic alternative to behavioral explanations. The SS retirement age policy implication is among the most direct applications of the income-mortality gradient to Social Security reform debates — an important connection to the DI literature. The key limitation: the paper does not pursue causal identification of why low-SES groups stagnate; it establishes measurement robustness, not mechanism.